Dose response for Listeria monocytogenes

Transcription

1 Dose response for Listeria monocytogenes IAFP Webinar Organized by: Microbial Modeling and Risk Analysis PDG All opinions and statements are those of the individual making the presentation and not necessarily the opinion or view of IAFP

2 Moderator Marcel Zwietering Wageningen University Audio is via your computer speakers Questions should be submitted via the Text Chat section at the bottom of the screen (can be done during the presentations, handled at the end).

3 Dose response for Listeria monocytogenes IAFP Webinar Organized by: Microbial Modeling and Risk Analysis PDG All opinions and statements are those of the individual making the presentation and not necessarily the opinion or view of IAFP

4 Contact information for presenters Dr. Régis Pouillot formerly Center for Food Safety and Applied Nutrition, Food and Drug Administration (USA) Prof. Dr. Fernando Pérez Rodríguez Department of Food Science and Technology, University of Córdoba (Spain)

5 Former visiting scientist US Food and Drug Administration Division of Risk and Decision Analysis

6 united included function tests infected L. monocytogenes rate consumption reported according safety statistical sensitivity corresponding campylobacter fig veterinary exposure methods distributions obtained prevalence herd months retail produce available impact Dose response for Listeria monocytogenes What did we learn from recent outbreaks? monocytogene potential brucellosis yaoundesize genotypes conditions infections period med times follow ing foodborne maximum salmonella infect contamination example cell temperature isolates estimation pregnant respectively treatment patients control method effect meat les detection national type table assessment genotype performed standard ml virus distributionestimates epidemic probability hiv prior specific test analysis mice due found data clinical ii water na results models storage salmon mean modeltime age considered variability based france hcv population food value total rates major css strains women days french diseases research study animals disease risk infection foods consumed level deli pasteur estimate significant dis animal infectious n transmission milk cameroon ci rte africa outbreaks positive using low result testedjournal growth estimated associated survey cells fresh values chicken strain listerialisteriosis health log hepatitis laboratory bacteria cfu samples observed w eeks microbiol public parameters countries products factors expected studies uncertainty bacterial lowerdifferences show n human duchildren described product sequences process contaminated incidence microbiology applied parameter approach including international reference quantitative sample compared concentration enterocolitica african species stage v accine bayesian Régis Pouillot, DVM, PhD IAFP MMRA PDG Webinar, June 1 st 2017

7 Listeria monocytogenes Few complete outbreak data are available to derive a dose-response Hispanic cheese outbreak in Los Angeles county, 1985 Listeria monocytogenes infection from butter in Finland for transplant patients, Lyytikäinen et al., J Infect Dis, 2000, Linnan et al., N Engl J Med, 1988, FDA/FSIS 2003, FAO/WHO 2004

8 Specific issues with listeriosis Dose Long incubation period Difficulties to find leftovers Bacterial growth Correspondence between the number of bacteria found in the sample and the actual quantity of ingested L. monocytogenes Heterogeneity in the level of contamination Did the cases ingested the most contaminated products? Response Heterogeneity in the response the underlying conditions of the consumers might be more important than the dose

9 Recent Data Collection in US Focus on data collection Celery Outbreak, TX, 2010 Cantaloupe Outbreak, Multistate, 2011 Caramel Apple, Multistate, 2015 Ice Cream, Multistate,

10 The case of Celery Celery Outbreak, TX, cases in inpatients of an hospital Data on underlying health issues Growth studies Transfer studies Limited prevalence and contamination level data Knudson Gaul, L et al., Clin Infect Dis, 2013, Sahu et al., Food Control, 2017; Kaminski et al., JFP, 2014

11 The case of Cantaloupe Jensen farms cantaloupe outbreak 147 cases, 30 deaths across 28 States No enumeration data USDA photo by Scott Bauer. Image Number K

12 The case of the Caramel Apples 35 illness including 11 associated with a pregnancy 3 meningitis among otherwise healthy children aged 5-15 Apple does not support growth (ph 3.2) Caramel does not support growth (low a w ) Caramel coated apple with a stick supports growth Up to 7 log 10 in few days at ambient temperature Actual level at time of consumption? Photo: el-apples-12-14/images/caramel-apple2-450px.jpg CDC website, K. A. Glass et al., mbio 6, e (2015). 12

13 The case of Ice-Cream 4 cases, linked to one product of factory A, observed in inpatients of a single hospital one additional case in the hospital, but the strain was not recovered from ice cream 5 cases linked to a second factory (factory B) over 5 years! CDC website 13

14 The case of Ice-Cream FDA collected more than 10 ½ pallets of samples from the company s Factory A FDA optimized enumeration methods for the contamination in these products using both MPN and direct plating methods The MPN method used a 3 10 g, g, g and g dilution scheme ( ) FDA collected sales data With address Categorization Hospitals, Schools, Chen et al., JFP, 79(11) 2016; Chen et al., IJFM, 241,

15 Jan 2010: First case linked to the brand Sale log starting date: Nov 7 th, 2013 Timeline Jan 2014: First case linked to the factory A: Patient #1 Product B available, tested positive: May 21 st, 2014 Product A available, tested positive: Nov 6 th, 2014 Mar 2014: Second case linked to the factory A: Patient #2 Oct 2014: Third case linked to the factory A: Patient #3 Pouillot et al., EID 22(12) 2016 Product C available, tested positive: Dec 8 th, 2014 Cleaning and overhauling of the production line: Jan 2015 Jan 2015: Fourth case linked to the factory A: Patient #4 Mar 13 th, 2015: Products removed from the market

16 The 4 cases Onset ranges from January 2014 through January 2015 (over one year) All 4 were >67 and < 84 years of age All 4 had underlying conditions that contributed to compromised immune function before exposure Ate the product via milk shakes Two patients had two milkshakes One patient had three milkshakes (unrecorded data for the fourth one) Pouillot et al., EID 22(12) 2016

17 Frequency Frequency Frequency Enumeration data Product A, from factory A 2,290 samples of Product A tested all but 13 samples were positive (99.4% positive) Highly consistent low contamination levels 58% below 5 MPN/g 77% below 10 MPN/g 92% below 20 MPN/g 98% below 50 MPN/g 99.8% below 100 MPN/g MPN/g 4 samples > 100 MPN/g* one >208 MPN/g (direct plating: 357 cfu/g), two = 208 MPN/g (direct plating: 142 cfu/g and Non Available), one 139 MPN/g (direct plating: 177cfu/g) Chen et al., J Food Prot 79: 1828: MPN/g log10(mpn/g)

18 Enumeration of L. monocytogenes in Different Lots of Product A Chen et al., J Food Prot 79: 1828:45 Cleaning and overhauling of the production line 18

19 4 cases in the same hospital? How can it be? Raises the concern of a systematic problem within the hospital BUT It appears that this hospital bought 55% of Product A sold to hospitals The probability to have the 4 cases in this particular hospital was not that low (9%, actually) Pouillot et al., EID 22(12) 2016

20 Fn(x) Under the assumption of no growth, the probability to have a high level of contamination of the product A was low But not null Product A Milk Shake Growth in milk shake: limited (Chen et al., 2016) Contamination of the milk shake machine: possible (the machine was inoculated multiple times, every days, during a long period of time), but not detected Answer: probably yes but impossible to say. Did the cases eat a low number of bacteria? ScoopsMC Pouillot et al., EID 22(12) 2016

21 Did some individuals and susceptible individuals eat contaminated products and did not get sick? Best selling ice cream brand in the US in 2014*, 100% contaminated products during months / years Our estimates Millions of contaminated servings sold to the population Tens of thousands contaminated servings sold to pregnant women: no case identified Thousands contaminated servings sold to highly susceptible population 4 identified cases Answer: Yes, a lot! * source: wikipedia

22 The tip of the ice cream Four cases: dose? Millions of non-cases

23 Back to the dose response What is actually needed to derive a doseresponse from outbreak data Ingested Dose Outcome (sick, not sick)? Do we need the actual dose for each consumer? No Exponential dose response Prob(infection dose) = 1 exp(- r dose) r subpopulation nb of nb of cases inthe subpopulation L. monocytogenes ingestedby the subpopulation See supplemental material, Pouillot et al., EID, 22(12) 2113:9

24 Number of cases in the population 4 in the highly susceptible population, observed Number of L. monocytogenes ingested by the population = number of servings in the population average number of L. monocytogenes per serving number of servings: function of the starting date of the contamination: at least one year, maybe 5 or more? Average number of L. monocytogenes per serving Robust estimation from the thousands of tested samples (if we assume that the least months are representative of the whole episode) if no growth

25 Pouillot et al., EID, 22(12) 2113:9 Results Scenario model Whole population (4 cases) Highly susceptible (4 cases) Pregnant (0 case) Age > 75 y (4 cases) Lower Nb of ingested cells r parameter < Upper < r parameter < Compare to FAO/WHO : r = from the Finnish butter outbreak r = from the Hispanic cheese outbreak r = from epidemiological data

26 Probability of Adverse Effect 1.E+00 1.E-01 1.E-02 1.E-03 1.E-04 1.E-05 1.E-06 1.E-07 1.E-08 1.E-09 1.E-10 1.E-11 1.E-12 1.E-13 1.E-14 1.E-15 Dose (log 10 cfu/serving) Butter Ice cream, All populations, lower estimate Ice cream, Highly Susceptible population, lower estimate FAO/WHO Hispanic-style Cheese Ice cream, All populations, higher estimate Ice cream, Highly Susceptible population, higher estimate HHS/USDA

27 Conclusions We ll probably never get better data Impossible to know for sure what was the ingested dose of specific individuals Not needed to derive a dose response model! Paradox: more information on the non-cases than on the cases. Corresponding dose-response model comparable to previous doseresponse models developed from outbreaks Other outbreaks: limited diffusion of products contaminated at relatively high level Here: large distribution of products contaminated at low level Probability of infection following the ingestion of a given dose higher than for models derived from epidemiological data Bias when working on outbreaks: what about large diffusion of contaminated products that never lead to a recorded outbreak? 27

28 Keys for the Risk managers A no-growth product, with a very low average level of contamination (8 cfu/g) did cause an outbreak Directly OR indirectly It is likely that most patients were exposed to ice cream with < 100 cfu/g A high dose can t be excluded inoculation of the shaker at each serving. Biofilm? Growth? The large distribution of a no-growth product, with a low average level of contamination didn t cause a massive outbreak The underlying health of the patient, cell-mediated immune status, medications and repeated exposure (?) may be more important than the dose Sufficient for risk management? Few products that support growth (e.g. cheese) Outbreak Very large distribution of a contaminated product that does not support growth (e.g. ice-cream) Outbreak (directly or not) 28

29 Thank you Co-authors of the manuscripts and everyone involved in the data collection for these outbreaks Among other papers

30

31 Selection of a Listeria monocytogenes dose-response model for risk assessment in ready-to-eat products Dr. Prof. Fernando Perez Rodriguez Department of Food Science and Technology University of Córdoba (Spain)

32 Outline 1. Risk assessment scheme 2. D-R models: types and approaches 3. Review of Listeria D-R models 4. Selection of Listeria D-R models for QMRA 5. D-R model integration into QMRA

33 Quantitative Microbial risk assessment EXPOSURE ASSESSMENT HAZARD CHARACTERIZATION RISK CHARACTERIZATION

34 Dose-response (D-R) model Mathematical function that may be used to describe the relationship between dose and the magnitude of a response on a continuous scale in an individual. Adapted from Haas et al. 1999

35 Approaches for D-R modelling Black box models: f(p x C x Freq x Serv) = Surveillance knowledge-based model: the extent and severity of the disease as a result of the ingestion of cells by an individual or a population is known. Mixture model: in some cases, a mixture of the previous two models is created, e.g. a surrogate animal model calibrated with the actual incidence of the disease in a human population.

36 CDF Black box models: D-R modelling Empirical models are extensively applied in risk assessment, assuming a distribution of population tolerance: Limited in case of extrapolation Cumulative distribution Surveillance = f(p, C, Freq, Serv) Censored data

37 D-R models Mechanistic models are potentially more flexible: on a set of biologically plausible, mechanistic assumptions. They should account for the host-pathogen-food interaction (Hoelzer et al., 2013) Mechanistic models are based on the Independence of Action of microorganisms: No-theshold mechanistic models are the most frequent models

38 Next-generation microbial D-R models Key-event dose-response models (Buchanan et al. 2009) Buchanan, R. L., Havelaar, A. H., Smith, M. A., Whiting, R. C., & Julien, E. (2009). The Key Events Dose-Response Framework: its potential for application to foodborne pathogenic microorganisms. Critical Reviews in Food Science and Nutrition, 49(8),

39 Listeriosis L. monocytogenes is a psychrotropic microorganism able to produce a foodborne diseases Listeriosis is mostly related to relatively high doses and Elderly population (>64) is most affected group, particularly >84 year (ECDC/EFSA, 2016) In 2014: EU case fatality was 17.7% among the 1,524 confirmed cases with known outcome

40 D-R models for L. monocytogenes Pathogen Host Food Matrix Virulence of strains with/without PMSC in inia. Outbreak data, species differences in pathophysiology, guinea pigs and rhesus monkey dose-response data Prevalence and concentration of strains with/without PMSC in inla Determinants of virulence variability among strains: fixed genetic determinant, transient determinant (e.g. stress response) Differential gene expression as a function of the environment (e.g. food) Improved outbreak data New animal models Better understanding of determinant of susceptibility Alternative models Food implicated in outbreaks; characterization of contamination patterns, representativeness and accuracy of count data D-R data for strains with varying virulence Association of strains with clinical manifestation (e.g. meningitis) Growth and concentration of L. monocytogenes in the intestinal lumen. Impact of L. monocytogenes gene expression at each step of infection Better understanding of L. monocytogenes pathophysiology (role microbiota) Better understanding of host susceptibility Food characteristics modulating L. monocytogenes gene expression Food characteristics related to listeriosis and listeria growth Identifying infective dose and concentration in outbreaks. based on the Dose-Response Workshop outcomes (adapted from Hoelzer et al. (2013)). Available Data need Long-term

41 Review of D-R models for L. monocytogenes Differences in the population groups Low-risk group vs. High risk group Age-based groups: Intermediate-age*, Elderly (over 60-65) and Pregnant Elderly, pregnant and immunocompromised population Different r-values (point-estimates) Mostly USA-originated D-R models Endpoint: 80% illness, infection (animals) and death

42 Review of D-R models for L. monocytogenes Different mechanistic models in literature: Exponential dose-response model (77 %) Weibull-gamma model (12%) Specific dose-response model developed by internationally recognized institutions: FDA/FSIS (2003) WHO (2004) FDA/FSIS (2003) has been employed in 21 % risk assessments for L. monocytogenes WHO (2004) has been included in 32 % of studies.

43 D-R models for L. monocytogenes The exponential form of the DR model relationship is mostly preferred due to Mechanistic, it does not present threshold, it is a one-parameter model, and it has been widely used by different organizations including FAO/WHO and FDA/FSIS Exponential dose-response model P (ill; d, r) = 1-exp(-rd) where ill stands for illness, d refers to dose, and r is the probability of developing listeriosis from the ingestion of one bacteria cell in a given specific serving.

44 Selection of D-R models for risk assessment Tool to evaluate the quality of the Exponential dose-response models currently available: Application of Numeral Unit Spread Assessment Pedigree (NUSAP) system The NUSAP system (Boone et al., 2009) is intended to assess data quality resulting from uncertainties that are hard to quantify such as methodological and epistemological uncertainties, and that are not systematically taken into account in scientific studies. NUSAP scoring system Objective scores Assessors Weights to Pedigree Criteria Experts

45 Selection of D-R models for risk assessment Pedigree criteria Proxy: Year of publication of the dose-response model. Geographical origin of primary data. Not applicable for animal models. Empirical basis: Primary source of data. Number of independent sources for the primary source of data. Number of subpopulation groups from which data were analysed. Methodological rigor Inclusion of variability and uncertainty. Statistical analysis. Not applicable for Buchanan et al. (1997) approach. Number and descriptions of endpoints. Publication source. Validation: Validation of the dose-response model with other datasets.

46 Selection of D-R models for risk assessment Self-assessment Pedigree criteria weight weight Assessor Score system Expert Objective scoring Scoring Final score

47 Selection of D-R models for risk assessment Example with proxy : PEDIGREE CRITERIA SCORES Score: 4 Score: 3 Score: 2 Proxy Time Space Exact measure of the desired quantity (e.g., measurements from the same geographically representative area as that being investigated) Data from the last 5 years (measured, if Data from more than 1 not available publication date). European country. Good fit or measure (e.g., measurements used from another geographical area but representative) Data from the last 10 years (measured, Data from 1 European country. if not available publication date). Well correlated but not measuring the same thing (e.g., large geographical differences, less representative) More than 10 years old study. Data from US, Canada or NZ. Score: 1 Weak correlation (e.g., very large geographical differences, low representativeness) More than 20 years old. Data from other countries.

48 Selection of D-R models for risk assessment Outcome: Arithmetic versus geometric sequence (arithmetic sequence, i.e. 1, 2, 3 and 4). General agreement in the difficulty of FDA/FSIS model to be implemented; the model is neither readily reproduced nor readily defined the use of two dose-response models: Pouillot et al. (2015), representing a novel approach to describe L. monocytogenes doseresponse relationship; and FAO/WHO (2004), an institutional approach internationally recognized and easy to reproduce.

49 D-R models into risk assessment Prevalence Concentration Retailing Recontamination Non-prevalent units (%) Prevalent Storage time & temperature Transportation Products formulation Home storage Consumption Serving size, frequency & level of consumption Growth model Time-temperature Final concentration Serving size Serving number Dose response models Zero Final doses Probability illness/number of cases

50 D-R models into risk assessment Individual risk: Population risk: Probability distribution for probability of illness from a single hamburger meal predicted by the E. coli O157:H7 Process Risk (Adapted from Cassin et al. 1998) Pill i = Prevalence i x Dose i x r-value Integration of probability distribution for the probability of illness for the whole population (total number of exposures) n i=1 Pill i

51 D-R models into risk assessment

52 D-R models into risk assessment -Risk model for listeriosis in EU Elderly from Cooked meat-

53 D-R models into risk assessment -Introducing D-R model parameters from FAO/WHO (2004)-

54 D-R models into risk assessment -Initial concentration at retail-

55 D-R models into risk assessment -log increase from retail to home-

56 D-R models into risk assessment -Serving size-

57 D-R models into risk assessment -Exposure Assessment- Hazard Characterization-

58 D-R models into risk assessment -Simulation output- Individual risk 3.53E09 servings in EU Elderly population risk

59 Take-home message Different approaches for modelling Dose-response (D-R): Single hit without threshold as a pseudo-mechanistic model The Exponential model mostly used for describing Listeriosis D-R relationship NUSAP scoring system as tool to guide D-R model selection in risk assessment NUSAP High scores for FAO/WHO (2004) and Pouillot et al. (2015) The D-R integration into QMRA can be intended to estimate individual and population risk Easy-to-use software can be employed to implement D-R models in an stochastic environment

60 Questions

61 Contact information for presenters Dr. Régis Pouillot Prof. Dr. Fernando Pérez Rodríguez Marcel Zwietering, PhD Keep your browsers open to complete the survey!!